Improved Alglorithem And Reasearch Of License Plate Recognition System

In order to solve the problem of accurate location of license plate,Chinese character segmentation and Chinese character recognition,an improved license plate location algorithm and Chinese character based on special character position are proposed in this paper.Cutting algorithm and different location character use different network model to carry out the character recognition algorithm,and use the time complexity,space complexity and many contrast experiments to verify the rationality and accuracy of the algorithm proposed in this paper,which can effectively solve the above problems.First,this paper uses the mean fuzzy algorithm of multiple iterations to smooth the image,remove a lot of noise pollution and save time for the further positioning operation.After the preprocessing of the algorithm,the improved AdaBoost algorithm is used to carry out the license plate location operation,and the LBP feature is used instead of the traditional HOG feature and Haar.It is proved that the performance of the mean fuzzy algorithm using multiple iterations is better than the traditional Gauss fuzzy algorithm and the improved algorithm using LBP feature extraction is superior to the traditional feature.After the positioning,the improved algorithm of size judgment and angle judgment algorithm is used to deal with the tilt phenomenon.Secondly,in the segmentation module of the license plate character,the Chinese character segmentation algorithm based on the "special location" is used to obtain the position of Chinese characters,and it has a clear and accurate outside rectangle.It can accurately segment Chinese characters,and there will not exist the fracture of the head of Chinese characters,the character adhesion,and the influence of characters.The effect of segmentation and the improved algorithm are applied to deal with the influence of the decoration such as redundant slot and rivet on character segmentation.Finally,the characters are identified by different network models.The first character in the license plate uses the SVM algorithm to train the Chinese character network model.The second characters use the BP neural network algorithm to train the alphabet network model,and the remaining characters use the alphabet and digital mixed network.Through three experiments,the improved character recognition algorithm is better than the traditional neural network algorithm and SVM algorithm,and the overall positioning rate,recognition time and Chinese character accuracy are tested.All of these algorithms are superior to the traditional algorithms,and have high practical and social value.